“The most successful drug discovery project is the one with the fewest compounds synthesised”.
This is a statement I have made many times over the course of my career, with 28 years as part of multiple drug discovery engines at major Pharma, specialised DMPK and wider-service Contract Research Organisations (CROs) and now a DMPK consultancy company. Let’s dive into what I mean by this statement.
Everyone who is part of the small-molecule Drug Discovery model is now familiar with the Design-Make-Test-Analyse (DMTA1) cycle and practices it in their attempts to synthesise a clinical candidate that has the potential to become a novel drug to treat human diseases. DMTA usually starts with a hit, lead or fast-follower compound that has moderate potency at the desired target. Derivates based on this original hit are then submitted for further in vitro pharmacology and DMPK/Physicochemical testing, often in a parallel Tier 1 screen.
For a compound that will be delivered orally, and where the liver is the major metabolising/clearance organ, the key tier 1 DMPK/Phys.Chem assays are kinetic solubility, logD7.4 and intrinsic clearance in either liver microsomes or hepatocytes of the relevant pre-clinical species and human.
Understanding unbound intrinsic clearance (CLint,u ) early in a project is key as it is directly proportional to unbound exposure (AUCu), assuming fraction absorbed and fraction of drug escaping intestinal metabolism are high. Therefore, focusing on decreasing the CLint,u is a key strategic lever in maximising unbound blood exposure from a given oral dose.
Therefore, one can see that both potency data from the in vitro pharmacology department and the key DMPK data must be reviewed in parallel to understand the Structure-Activity Relationship (SAR) and design better compounds for the next cycle.
The iterative DMTA cycle is the key to success, assuming that potency and DMPK are not mutually exclusive and can be optimized in parallel i.e. one cannot only design potent but metabolically unstable compounds or non-potent but metabolically stable compounds.
It is then logical to think that if the DMTA cycle was made faster, more compounds will be made and tested for the budgeted time-period leading to better success. Typically, drug discovery teams want to push for a 5 working day or a 1-week DMTA cycle time.
In many cases, this has led to integrating artificial intelligence (AI) into DMTA2,3 cycles. Worryingly, this often ignores DMPK/ADME considerations altogether, or focus solely on using predicted DMPK parameters4 rather than measured.
However, even without a reliance on AI, there are several potential issues with this need for speed. To allow for effective DMTA cycles, the pharmacology and DMPK/Phys.Chem. data from the current round must be fully understood to design better compounds ready for the next round of synthesis. Arguably, 5-days is not enough time to fully prosecute the SAR data and then synthesise the next wave of improved compounds. In addition, it is still common practice for medicinal chemists to have a target number of compounds to be synthesised as part of their personal objectives. When faced with this kind of pressure, why would one prioritize good science and make the more relevant compounds if they are more synthetically challenging and time-consuming? One can already see a potential issue brewing.
From the DMPK side, to meet the 5-day DMTA cycles, assays are often cut down to the bare bones and run only in a fully automated manner. It is common for assays to have only one replicate, limited time‑points, or to be run only in one standardised format, allowing no flexibility to better answer project questions e.g. by altering substrate concentration, cell number, time-points, replicates or investigating active uptake etc. For a Biotech company with limited financial resources this also often means outsourcing to the cheaper CROs in order to fit as many DMTA cycles as possible into the allocated aggressive timelines and budget.
Over the past 20 years of outsourcing to various global CROs, and being the founder of a UK-based DMPK specific CRO for 9 of those 20 years, I have regularly witnessed incorrect DMPK data. Sometimes, there has been a focus on a flawed DMPK screening cascade to drive the project forward (see ‘What do you mean I don’t always need to screen in hepatocytes?!’). On other occasions, data from in vitro assays or in vivo PK studies from 3rd party providers has been erroneous. Over those 20 years this has arguably got worse, which is very concerning.
One question that regularly occurs from our customers is “We have no in vitro – in vivo correlation (IVIV-C), and we regularly under‑predict in vivo clearance – how do we optimise compounds?”. It often requires a seasoned DMPK specialist to know when data is wrong or if the incorrect screening cascade has been adopted, and most often these DMPK experts are not employed in Biotechs. For the IVIV-C example above, we often find that once assays are repeated at a different CRO, with an improved assay design, and/or with further supporting follow-on bespoke DMPK assays, there is actually a good IVIV-C, and we do have a way forward to help them optimise compounds from a DMPK perspective.
Why, therefore, are drug discovery projects obsessed with the 1-week DMTA cycle? Wouldn’t a 2-week DMTA cycle actually lead to better quality candidate drugs via more accurate, reliable and predictive DMPK screens? Would it not be better to allow enough time to fully evaluate the SAR and give the medicinal chemists more flexibility to synthesise the next logical compounds, even if they are more complex and difficult to make? Making this change would allow CROs to be more flexible and offer higher quality data rather than focusing on the most cost-effective and quickest studies.
“We make ’em cheap and fast. Not good, just cheap and fast.” Burt Reynolds, 1982
2 Augmenting DMTA using predictive AI modelling at AstraZeneca
3 Overcoming DMTA Cycle Challenges: A Unified AI-Driven System for Efficient Drug Design